chem Schema

• Primary Key: (source, id)
• Tables: 1

The chem schema provides a unified view of chemical compounds from both the cc (Chemical Component Dictionary) and prd (BIRD Reference Dictionary) schemas. This enables cross-schema chemical searches without needing to query each source separately.

Why This Table Exists

The prd.brief_summary table only has canonical_smiles for PRD entries that have their own PRDCC file (~802 of ~1175 entries). The remaining ~373 entries are "single molecule" PRDs whose structure is defined by a CCD component -- they don't have PRDCC files, so prd.brief_summary.canonical_smiles is NULL. Their SMILES exist in cc.brief_summary instead.

The chem.compounds table solves this by combining all compounds from both sources into one searchable table. You don't need to worry about which source has the SMILES -- just query chem.compounds.

See PRD SMILES Coverage for details.

How It Works

The chem.compounds table is populated by the pmb compounds command, which:

1. Extracts compounds from cc.brief_summary (~50k CCD entries)
2. Extracts compounds from prd.brief_summary (~802 BIRD entries with SMILES)
3. Combines them into a single table with a source column ('cc' or 'prd') to distinguish origin

The table is also automatically refreshed after pmb update when the cc or prd pipelines run.

tip

"Single molecule" PRDs are already included as source = 'cc' entries (via their chem_comp_id), so all PRD-related compounds are searchable in this table.

RDKit Integration

Like the cc and prd schemas, the chem schema has full RDKit support:

1. mol column -- stores RDKit molecule objects generated from canonical SMILES
2. GiST index on mol for fast substructure and similarity searches
3. RDKit descriptor columns -- molecular weight, LogP, TPSA, HBA, HBD, rotatable bonds, rings, formula
4. Chemical search SQL functions:
   • chem.similar_compounds(smiles, threshold) -- Tanimoto similarity search across all sources
   • chem.substructure_search(smarts) -- substructure matching across all sources

Example Queries

-- Find all compounds (cc + prd) similar to aspirin
SELECT * FROM chem.similar_compounds('CC(=O)Oc1ccccc1C(O)=O', 0.5);

-- Substructure search across all sources
SELECT id, source, name, canonical_smiles
FROM chem.compounds
WHERE mol @> 'c1ccccc1'::mol;

-- Compare compound counts by source
SELECT source, COUNT(*) FROM chem.compounds GROUP BY source;

-- Find PRD compounds that share a CCD component
SELECT id, name, cc_comp_ids
FROM chem.compounds
WHERE source = 'prd' AND cc_comp_ids IS NOT NULL;

compounds

Column	Type	Description
id	text	Compound identifier (comp_id for cc, prd_id for prd)
source	text	Source schema: 'cc' or 'prd' (CHECK constraint enforced)
canonical_smiles	text	Canonical SMILES string
name	text	Compound name
formula	text	Molecular formula
cc_comp_ids	text[]	Associated CCD comp_ids (self-referential for cc, linked chem_comp_id for prd)

note

The mol column and RDKit descriptor columns (rdkit_mw, rdkit_logp, rdkit_tpsa, rdkit_hba, rdkit_hbd, rdkit_rotbonds, rdkit_rings, rdkit_formula) are added automatically by pmb setup-rdkit and are not shown in the table above.